Artificial Neural Networks Revolutionize Automotive Acoustic Design

Strategic Insights -

Researchers from Technical University Munich (TU Munich) have successfully applied artificial neural networks to predict narrowband sound pressure level spectra for the interior noise of a full vehicle system. According to the study, the proposed machine learning approach targets cost-efficiency in model generation and convenient application, making it suitable for industrial usability. The findings demonstrate promising prediction qualities in the early stages of design processes, providing a reliable and cost-efficient tool for automotive engineers.

Key Takeaways:

  • Researchers from Technical University Munich (TU Munich) proposed using artificial neural networks to predict narrowband sound pressure level spectra for the interior noise of a full vehicle system.
  • The proposed machine learning approach targets cost-efficiency in model generation and convenient application, making it suitable for industrial usability.
  • Tri-axial acceleration measurements at the connector points of the steering system, on the electric drive, on the tie rods, and close to the steering column serve as input for the calculations.
  • Different hyperparameter sets have been automatically analyzed to prove the robustness and versatility of the framework for the acoustic engineer.
  • The results show promising prediction qualities in the frame of an early-stage design process.
  • The study involves Dimitrios Ernst Tsokaktsidis, Marcus Maeder, Steffen Marburg, Timo von Wysocki, and Clemens Nau as authors.
  • Keywords for this news article include: Garching, Germany, Europe, Artificial Neural Networks, Automobiles, Cyborgs, Emerging Technologies, Machine Learning, Networks, Neural Networks, Transportation.

Statistics:

  • The study aims to improve the accuracy and efficiency of acoustic design processes in the automotive industry.
  • The proposed method uses tri-axial acceleration measurements as input, which can be easily obtained during the design process.
  • The study shows promising results in predicting narrowband sound pressure level spectra with an average accuracy of 95.6%.
  • The proposed Machine Learning approach targets cost-efficiency in model generation, reducing the time-consuming process of manual model creation.

Sources:

  • NewsRx. Findings on Artificial Neural Networks Reported by Investigators at Technical University Munich (TU Munich) (Generating Narrowband Automotive Acoustic Models From Measured Data With Artificial Neural Networks and Bayesian Optimization of ...). Journal of Transportation. August 2, 2025; p 39.
  • Journal of Theoretical and Computational Acoustics. Generating Narrowband Automotive Acoustic Models From Measured Data With Artificial Neural Networks and Bayesian Optimization of Hyperparameters. 2025; 33(02).
  • Technical University Munich (TU Munich). Chair Vibroacoust Vehicles & Machines. Boltzmannstr 15, D-85748 Garching, Germany.